METHODS AND SYSTEMS FOR UNOBTRUSIVE SEARCH RELEVANCE FEEDBACK
Methods and systems for receiving feedback from a user regarding the relevance of search results to a search query are provided. The method includes presenting at least one search result responsive to a search query by the user, selecting ones of the at least one search result that are determined by the user to be relevant to the search query, and increasing a relevance ranking of the selected search results in future searches based on the determination.
Embodiments of the disclosure relate generally to methods and systems to facilitate information searching and more particularly, to methods and systems for determining a relevance feedback for search results unobtrusively.
At least some known search methodologies and search engines incorporate interactive relevance feedback as a method to improve the relevance of search results. However, interactive relevance feedback is difficult to obtain from users that are more interested in the search at hand than supplying interactive relevance feedback. Users are also reluctant to provide interactive relevance feedback due to the extra demand of time to provide the feedback and the distraction away from the search activity. In some cases, users indicate that providing interactive relevance feedback may affect the search result in an unpredictable way.
Other known systems incorporate the relevance feedback system into the display of the search results. Incremental feedback is a method used in search systems to seamlessly integrate feedback tasks. The users are shown only one document at a time. Users are requested to assess the relevance of the document before being presented another document. Some systems cluster the search results and effectively derive user's search intent without interrupting user search task. These search systems cluster the documents in several topics, and adjust the search terms according to the user's selection of clusters. In this method, the user does not need to explicitly identify their intentions, and their browsing behavior of traversing the document clusters indicates their interest. However, such incremental feedback is inefficient in browsing the results—only one document is presented each time. Also, incremental feedback may perform well when retrieving documents that have very similar contents but does not perform well in retrieving documents that cover several topics. The clustering method only helps to identify the topics the user is interested in, but provides no feedback information for a particular document.
What are needed are methods and systems for search support that can make searches more efficient for responding to requests for information by using knowledge gained unobtrusively by user feedback concerning the relevance of each document in the search results.
SUMMARYIn one embodiment, a method of receiving feedback from a user regarding the relevance of search results to a search query includes presenting at least one search result responsive to a search query by the user, selecting ones of the at least one search result that are determined by the user to be relevant to the search query, and increasing a relevance ranking of the selected search results in future searches based on the determination.
In another embodiment, a system for improving the relevancy of electronic search results over a plurality of searches includes an adaptive search engine communicatively coupled to a plurality of data items from one or more data sources stored in at least one database, the data sources searchable using a search query entered by a user wherein the search query includes at least one keyword. The system also includes a user interface configured to display to the user a ranked search result listing of data items in a first display area, the user interface including a data item repository area wherein when a data item listed in the first display area is selected and copied to the data repository a relevance of the data item with respect to the at least one keyword is increased.
In yet another embodiment, a method of performing searches of a plurality of data sources that each includes at least one data item wherein the searches are performed using relevance feedback from previous searches. The method includes presenting a first search result of the data sources based on a search query, wherein the search result includes at least one data item in ranked order based on a relevance of the search item to the search query, unobtrusively monitoring a selection of data items by a user, associating selected data items with an increased relevance to the search query, and presenting a second search result wherein the ranking of the data items included in the search result is based on the association from a previous search.
The disclosure describes a document repository for document collection and a note panel for copying and pasting texts and images from returned documents from a query. Users' document collecting action and note-taking action is used as the base of relevance feedback. The relevance feedback is collected unobtrusively but effectively, and greatly improves the search relevancy. This unobtrusive interactive relevance feedback method overcomes the drawbacks of traditional interactive relevance feedback method—the extra time needed and interruption of tasks. It provides a novel, efficient and precise way of collecting interactive relevance feedback. Different with prior relevance feedback methods for search engine, no extra time is required to fill a feedback form. More useful feedback information is collected from search users without users' notice. Search results are improved. This disclosure describes novel unobtrusive methods and systems to collect search result relevance feedback from a user. Such feedback is then used in relevance feedback algorithms during subsequent search queries to improve the search efficiency and accuracy.
The methods and systems include a search document collection panel for the user to collect user interested documents and a note panel for the user to take notes. The user can drag and drop documents from the document collection panel for further reference. The user can also open a document and copy and paste text, audio, video, and images from the document to be saved in the notes.
A document repository is designed for user to drag and drop documents determined to be useful after the search results returned from a query. When the user drags and drops a document, a positive assessment is marked for the document for the query. A note panel is designed for the user to copy and paste text and other objects such as video, audio, and image clips determined to be useful When the user copies and pastes from a document to the note panel, a positive assessment is marked for the document for the search. A document and a query will be highly correlated if the document receives some positive assessment from the user for the query. And this information will be used to return the same document and rank the document higher when the user performs subsequent searches.
The methods and systems describe a method of receiving relevance feedback from the user without interrupting the user's search activities. The user feedbacks facilitate improving the search performance for a search engine in many different ways, such as improving search relevancy in future searches and building a user profile such that the search results are more customized for the user.
Repository database 28 includes product data such as, without limitation, design drawings, maintenance records, technical manuals, and regulatory directives for systems, components, and parts of for example, an aircraft. Repository database 28 further includes project data such as a project name, an origination date of the project, a status of the project, a model number to a product that the project pertains, an Air Transport Association (ATA) System Identification chapter number, and a customer identification number for the project. Repository database 28 may also include models such as CAD models, layouts, engineering drawings, and product structure data. Repository database 28 includes commercial information such as supplier and consumer information for various parts and products. Repository database 28 includes images, multi-media and product commissioning and in-service test results.
Computers 56 are a part of a Technical Data Search Engine system (TDSE), such as each of TDSE systems 58. As an example, computers 56 of user interface site 26 are located within an Operations Center or Service Engineering Group. In the example, computers 38, 40, 42, 44, and 46 of user interface site 24 are located within the engineering divisions and used by engineers with permissions for access to system 10. Server site 12 includes an application server 60, a database server 62, a mail server 64, and a directory server 66. Any of servers 54, 60, 64, and 66 are connected to a data repository 68 via database server 62. The repository data stored within data repository 68 is the same as the repository data stored within repository database 28.
Any of servers 54, 60, 62, 64, and 66 execute the methods for a collaborative task-based search model. The methods for the task-based search model include performing document management functions, which include searching for and selecting the repository data from data repository 68, creating the repository data for optional storage into data repository 68, editing the repository data, and retrieving the repository data from data repository 68. Moreover, the document management functions include linking the repository data within data repository 68, and organizing the data in accordance with naming conventions received from users searching for information relating to one or more customer service requests.
The project management functions also include, but are not limited to, receiving service requests directly from customers or from the customer through input by a user, receiving search criteria from the user to initiate a search of data sources selected by the user, generating a project folder to store search results that are determined to be relevant to the service request by the user. Additionally, the user may select portions of documents to insert into a note area of the database and associated with the project folder to further specify the relevance of certain search results. The project folder is accessible to other users who may receive similar service requests in the future so that previous searching may be reused to facilitate responding to the subsequent service request in a shorter amount of time than the first service request.
The TDSE application is organized around the concept of a project. The user creates a project in response to a particular search interest for example, a service request for a customer. A user can also initiate searches or generate notes. From the search results, a user can associate reference materials with a project. Queries and Notes are referred to as project actions. Such Queries and Notes are saved and can be recalled for later review at any time.
The TDSE application is used to look up information from a variety of data sources so that a user can simultaneously and selectively access relevant documents. Based on a single user input, multiple searches are initiated. Results from diverse data sources are then aggregated, sorted, ranked, and presented consistently and coherently to the user.
A project may be added automatically by an incoming service request such as via an email from the customer. TDSE automatically receives and parses the data in the email that is entered by the customer. A project may also be generated manually by the user in response to a service request from the customer via for example, a telephone call. To add a project, the user selects a “Projects” tab 404 and selects a “New” button 406. The user then fills in data input fields presented on user interface 402. Specifically, the user enters a unique project name in an “SR” field 408 comprising alphanumeric characters. Optional fields may also be filled in with data that may be available at the time the project is originated or may be filled in at a later time when details of the service request are made clear. To fill in any optional fields, the user enters information in a “Name” field 410, a “Date” field 412, a “Status” field 414, a “Model” field 416, an Air Transport Association (ATA) System Identification chapter number in an “ATA” field 418, and a customer identification number for the request in a “Customer #” field 420. To initiate the addition process, the user selects an “Add” button 422.
To select or unselect data sources to be searched, the user selects a “Data Sources” button 516 and a data sources selection window 518 is displayed. The user determines which data sources 520 are used for the search by selecting a check box 522 associated with the determined data sources. The user selects a checked box to deselect the associated data source 520 to exclude data source 520 from the search. The user may also select a check box 524 corresponding to a desired scenario to select a group of data sources by scenario. Such scenarios include but are not limited to aircraft information, part information, troubleshooting, part interchangeability, part availability, and part information. The user then selects an “OK” button to exit data sources selection window 518. To initiate the search the user selects a “Search” button 528. The label on “Search” button 528 changes to “Stop Search” while the search is executing. When all the selected searches are completed, the label of “Search” button 528 changes back to “Search.”
The above-described methods and systems for unobtrusive relevance feedback collection from a user are cost-effective and highly reliable. The methods and systems include monitoring the selection of particular search results that are determined by the user to be relevant to the search being conducted and increasing the results rank in subsequent searches. Accordingly, the methods and systems facilitate producing relevance feedback for document searches in a cost-effective and reliable manner.
While embodiments of the disclosure have been described in terms of various specific embodiments, those skilled in the art will recognize that the embodiments of the disclosure can be practiced with modification within the spirit and scope of the claims.
Claims
1. A method of receiving feedback from a user regarding the relevance of search results to a search query, said method comprising:
- presenting at least one search result responsive to a search query by the user;
- selecting ones of the at least one search result that are determined by the user to be relevant to the search query; and
- increasing a relevance ranking of the selected search results in future searches based on the determination.
2. A method in accordance with claim 1 further comprising selecting one or more data sources to be searched from a plurality of data sources.
3. A method in accordance with claim 1 further comprising entering one or more keywords on which to search the selected data sources for search results relevant to the one or more keywords.
4. A method in accordance with claim 1 further comprising displaying a search document collection panel that includes a document repository and a notes panel.
5. A method in accordance with claim 1 wherein presenting at least one search result comprises presenting a ranked search result listing wherein the ranking is based on a determination of the relevance of the search result to the search query.
6. A method in accordance with claim 4 wherein presenting at least one search result comprises presenting a file identifier of at least one electronic file to the user in the search document collection panel.
7. A method in accordance with claim 6 wherein selecting ones of the at least one search result comprises dragging and dropping the file identifier from the search document collection panel to the document repository.
8. A method in accordance with claim 6 wherein selecting ones of the at least one search result comprises:
- opening the electronic file; and
- cutting at least a portion of the electronic file and pasting the portion to the notes panel.
9. A method in accordance with claim 6 wherein selecting ones of the at least one search result comprises entering a text based note into the notes panel.
10. A method in accordance with claim 9 wherein selecting ones of the at least one search result comprises dragging and dropping a note from the notes panel to the document repository.
11. A method in accordance with claim 9 wherein increasing a relevance ranking of the selected search results comprises increasing a relevance ranking of notes added to the document repository.
12. A system for improving the relevancy of electronic search results over a plurality of searches comprising:
- an adaptive search engine communicatively coupled to a plurality of data items from one or more data sources stored in at least one database, said data sources searchable using a search query entered by a user wherein said search query comprises at least one keyword; and
- a user interface configured to display to the user a ranked search result listing of data items in a first display area, said user interface comprising a data item repository area wherein when a data item listed in said first display area is selected and copied to the data repository a relevance of the data item with respect to the at least one keyword is increased.
13. A system in accordance with claim 12 wherein said user interface further comprises a note area, wherein when a note generated in said note area is selected and copied to the data repository a relevance of the note with respect to the at least one keyword is increased.
14. A system in accordance with claim 12 further comprising a data source selector configured to permit the user to select one or more data sources to be searched from a plurality of data sources.
15. A system in accordance with claim 12 configured to receive one or more keywords entered by the user on which to search the selected data sources for search results relevant to the one or more keywords.
16. A system in accordance with claim 12 configured to present a ranked search result listing wherein the ranking is based on a determination of the relevance of the search result to the search query.
17. A method of performing searches of a plurality of data sources that each includes at least one data item, the searches performed using relevance feedback from previous searches, said method comprising:
- presenting a first search result of the data sources based on a search query, wherein the search result includes at least one data item in ranked order based on a relevance of the search item to the search query;
- unobtrusively monitoring a selection of data items by a user;
- associating selected data items with an increased relevance to the search query; and
- presenting a second search result wherein the ranking of the data items included in the search result is based on the association from a previous search.
18. A method in accordance with claim 17 wherein the search query comprises at least one keyword.
19. A method in accordance with claim 17 wherein associating selected data items with an increased relevance to the search query comprises dragging and dropping the data item from the search result to a document repository.
20. A method in accordance with claim 17 wherein associating selected data items with an increased relevance to the search query comprises dragging and dropping a note from a notes panel to a document repository wherein a note includes at least one of a portion of a data item, an annotation entered by the user wherein the annotation includes a text entry, an audio entry, a video entry, and an image entry.
Type: Application
Filed: Mar 7, 2007
Publication Date: Sep 11, 2008
Patent Grant number: 8386478
Inventor: Yanxin Emily Wang (Sammamish, WA)
Application Number: 11/683,273
International Classification: G06F 17/30 (20060101);